On Mon, Mar 22, 2010 at 3:54 PM, David sheehan david.sheehanjr at
gmail.comwrote:
I don't have icc and icpc. I only have ifort and gnu compilers such g77,gcc
and g++. Since my application code works very well with ifort, can I build
up PETSC plus Hypre with ifort and gnu compilers? thanks.
Also, these are questions for petsc-maint. They are not about PETSc
development.
Matt
On Mon, Mar 22, 2010 at 3:58 PM, Satish Balay balay at mcs.anl.gov wrote:
try:
./configure --with-cc=gcc --with-cxx=g++ --with-fc=ifort --download-mpich=1
--download-hypre=1
Satish
On Mon, 22 Mar
MINRES-QLP: a Krylov subspace method for indefinite or singular symmetric
systems. (arXiv:1003.4042v1 [math.NA]) http://arxiv.org/abs/1003.4042
http://arxiv.org/abs/1003.4042Matt
--
What most experimenters take for granted before they begin their experiments
is infinitely more interesting
On Tue, Mar 23, 2010 at 11:31 AM, Satish Balay balay at mcs.anl.gov wrote:
On Tue, 23 Mar 2010, Jed Brown wrote:
On Tue, 23 Mar 2010 11:00:04 -0500, Keita Teranishi keita at cray.com
wrote:
I'd like to know if you have had any problem with 3.1.1 and if you
have any plan to introduce
That option dates from the time I was trying to use DA to do structured
finite elements. I have since given up on this, and use a CartesianMesh
which has more support for this. Of course, eventually they should merge.
Matt
On Wed, Mar 24, 2010 at 4:56 PM, Jed Brown jed at 59a2.org wrote:
I
On Wed, Mar 24, 2010 at 5:11 PM, Jed Brown jed at 59a2.org wrote:
On Wed, 24 Mar 2010 17:04:35 -0500, Barry Smith bsmith at mcs.anl.gov
wrote:
Sorry, I was so stunned when I saw that option I just automatically
nuked it.
I have pushed 3.1 and dev with the option again.
I usually \mathrm for stuff like this.
Matt
On Wed, Mar 24, 2010 at 6:13 PM, Satish Balay balay at mcs.anl.gov wrote:
Jed,
I get doc errors due to
http://petsc.cs.iit.edu/petsc/petsc-release-3.1/rev/34db3f4fbf0b
106] [107]
! Undefined control sequence.
l.4336dt^{n} = dt_{\text
On Thu, Mar 25, 2010 at 2:09 PM, Kai Germaschewski
kai.germaschewski at unh.edu wrote:
On Thu, Mar 25, 2010 at 12:00 PM, Lisandro Dalcin dalcinl at gmail.comwrote:
On 25 March 2010 01:14, Kai Germaschewski kai.germaschewski at unh.edu
wrote:
Fine with me (I'm the one who introduced it
On Thu, Mar 25, 2010 at 2:28 PM, Satish Balay balay at mcs.anl.gov wrote:
On Thu, 25 Mar 2010, Kai Germaschewski wrote:
On Thu, Mar 25, 2010 at 12:00 PM, Lisandro Dalcin dalcinl at gmail.com
wrote:
On 25 March 2010 01:14, Kai Germaschewski kai.germaschewski at unh.edu
wrote:
Fine
For those of you who care, the RSS updates from BitBucket are far superior
to those from our current site. I wonder if we can upgrade.
Matt
--
What most experimenters take for granted before they begin their experiments
is infinitely more interesting than any results to which their
Why would you attach an option instead of having an equiv API call? Strings
are ALWAYS a bad interface.
Matt
On Fri, Apr 2, 2010 at 9:29 PM, Dmitry Karpeev karpeev at mcs.anl.gov wrote:
On a somewhat related note: would it make sense to have the functionality
to
attach options or just
On Mon, Apr 5, 2010 at 8:47 PM, Lisandro Dalcin dalcinl at gmail.com wrote:
On 3 April 2010 23:32, Matthew Knepley knepley at gmail.com wrote:
Why would you attach an option instead of having an equiv API call?
Do PETSc have an equivalent API call for every option you can set
using
Why does this not match MatGetArray()? I know it looks good from MATDENSE,
but
shouldn't this be a special dense function? This is just annoying me since I
want to
pull this out without a stupid offset.
Matt
--
What most experimenters take for granted before they begin their experiments
is
Fixed.
Matt
On Thu, Apr 15, 2010 at 11:19 AM, Lisandro Dalcin dalcinl at gmail.com wrote:
While starting to update petsc4py to petsc-dev, I've found that (IMHO)
naming inconsistency... If appropriate, could any of you fix that for
me?
--
Lisandro Dalcin
---
CIMEC
On Wed, Sep 2, 2009 at 3:50 PM, Lisandro Dalcin dalcinl at gmail.com wrote:
libfast in: /u/dalcinl/Devel/PETSc/petsc-dev/include/ftn-auto
petschf.c: In function ?petscmemcpy_?:
petschf.c:42: warning: no return statement in function returning non-void
The problem here is that bfort hates
On Wed, Sep 2, 2009 at 4:18 PM, Lisandro Dalcin dalcinl at gmail.com wrote:
On Wed, Sep 2, 2009 at 6:02 PM, Matthew Knepleyknepley at gmail.com wrote:
On Wed, Sep 2, 2009 at 3:50 PM, Lisandro Dalcin dalcinl at gmail.com
wrote:
libfast in: /u/dalcinl/Devel/PETSc/petsc-dev/include/ftn-auto
as another bullet to
Supplementary Documentation, and I really have sent it to my
PETSC/Fortran colleagues.
Mark
On Sep 1, 2009, at 9:49 PM, Barry Smith wrote:
Beats me, that's why there are not many links to it.
Barry
On Sep 1, 2009, at 8:15 PM, Matthew Knepley wrote:
is a well
, but this can be argued about. My
rationale was
that it is an artifact of implementation that we have DA_1d, etc., and we
have ADDA
already (which is not yet converted).
Thanks,
Matt
On Sat, Feb 21, 2009 at 5:00 PM, Jed Brown jed at 59a2.org wrote:
On Sat 2009-02-21 15:45, Matthew Knepley wrote
1) Please don't cut off the stack. Its very useful when debugging problems
like this.
2) When reading one of these true/false options, it got an unprintable
character
instead of nothing (hence the weird error message). I recommend giving
a
valid 1/0 to every option. Its typically some
On Tue, Sep 22, 2009 at 12:02 PM, Lisandro Dalcin dalcinl at gmail.com wrote:
I commented about this in the past, but let's do it once more...
Just for reference, in petsc4py (accesible from core PETSc as TSPYTHON
type when configured --with-petsc4py) more or less uses Jed's model:
the user
That is Oliver's old code. It is really complicated. I believe even Oliver
thinks that any new implementation should start from scratch, using his
current PETSc code as a reference.
Matt
On Tue, Sep 22, 2009 at 1:15 PM, Dmitry Karpeev karpeev at mcs.anl.gov wrote:
I've found something that
beyond
a single linear algebra problem. Somehow, discretization needs to come in.
Matt
Dmitry.
On Tue, Sep 22, 2009 at 1:17 PM, Matthew Knepley knepley at gmail.comwrote:
That is Oliver's old code. It is really complicated. I believe even Oliver
thinks that any new implementation should start
I moved everything to a single module. This will introduce inevitable
errors. Just
report everything to me.
Thanks,
Matt
--
What most experimenters take for granted before they begin their experiments
is infinitely more interesting than any results to which their experiments
lead.
--
Its not automated. You add them to the fblaslapack (and the cblas would be
nice too). Then add a test for these methods in the BlasLapack.py configure
test (since some idiot impl will not have them). If its confusing, ask me or
Satish.
Thanks,
Matt
On Tue, Sep 29, 2009 at 12:10 PM, Jed
is a well-written page, but I don't think most people can find it. Where
else should we link it? FAQ? Documentation?
Matt
--
What most experimenters take for granted before they begin their experiments
is infinitely more interesting than any results to which their experiments
lead.
-- Norbert
I believe that the time reported there is collective sum of times divided by
the collective sum
of the stage times. If you look at the time imbalance, it is a staggering
9.7, which either means
1) The partition is really crap (which we know isn't true)
2) Some procs spend a lot of time
2009/5/5 Giuseppe Ghib? ghibo at mandriva.com
Hi.
I'm trying to package petsc 2.3.3, but I didn't understand the logic
of some options, in particular for umfpack. E.g. If I specify at the
config stage:
--with-umfpack-lib=[${_libdir}/libumfpack.a,${_libdir}/libamd.a]
and the ${_libdir}
On Wed, May 6, 2009 at 9:17 AM, Satish Balay balay at mcs.anl.gov wrote:
On Tue, 5 May 2009, Giuseppe Ghib? wrote:
Hi.
I'm trying to package petsc 2.3.3,
Why package this older version? Current version is 3.0.0.
but I didn't understand the logic of some options, in particular for
On Sat, May 9, 2009 at 1:09 PM, Barry Smith bsmith at mcs.anl.gov wrote:
After looking at the code I see why it ended up being so complicated. It
needs some thought.
The way the line search works is
1) if the full step satisfies the reduction criteria it returns
2) if the full step is
2009/5/11 Giuseppe Ghib? ghibo at mandriva.com
Satish Balay wrote:
On Tue, 5 May 2009, Giuseppe Ghib? wrote:
Hi.
I'm trying to package petsc 2.3.3,
Why package this older version? Current version is 3.0.0.
Because I've to use bundled with libmesh (http://libmesh.sourceforge.net)
I vote for kicking out BlockSolve95. It has not been maintained since 95. It
does parallel
ICC(0) and ILU(0). The ILU can be done by PILUT. It appears the only thing
we give up
is ICC(0) in parallel. IS this worth keeping this headache?
Matt
On Tue, May 19, 2009 at 1:25 PM, Satish Balay
On Tue, May 19, 2009 at 8:58 PM, Barry Smith bsmith at mcs.anl.gov wrote:
I concur, rip it out of petsc-dev
Okay, ripped.
Matt
Barry
Hmmm, maybe we should rip it out of petsc-3.0.0 but since it involves
lots of changes, h
On May 19, 2009, at 1:59 PM, Matthew Knepley
I think there is an inconsistency in our definitions. If you create a matrix
which
is 4x8 on 2 procs, then you end up with a row and column map like this:
rowMap colMap
[0..2) [2..4) [0..4) [4..8)
However, we use the column map for queries which then give nonsense. The
most
a symmetric partition, and
is
probably at the root of our problems with rectangular matrices.
Matt
On Tue, May 26, 2009 at 10:35 AM, Matthew Knepley knepley at gmail.com
wrote:
I think there is an inconsistency in our definitions. If you create a
matrix
which
is 4x8 on 2 procs, then you end
On Tue, May 26, 2009 at 12:11 PM, Lisandro Dalcin dalcinl at gmail.com wrote:
On Mon, May 25, 2009 at 10:20 AM, Jed Brown jed at 59a2.org wrote:
Is it really necessary that iscol be a sequential index set that covers
all columns, with csize selecting the local columns? I realize this
On Tue, May 26, 2009 at 12:18 PM, Jed Brown jed at 59a2.org wrote:
Lisandro Dalcin wrote:
The only real issue I see here is that the term diagonal block could
lead to misinterpretation. But the behaviour you commented is the one
I would expect.
The rest of this mail is perhaps
On Tue, May 26, 2009 at 12:29 PM, Jed Brown jed at 59a2.org wrote:
Matthew Knepley wrote:
It not computing Ab. Its computing Ab, A^2b, A^3b, etc.
Of course, but that's all possible if A*b is computable. Lisandro
demonstrated that A*x was fine, which it better be because of how x
On Tue, May 26, 2009 at 1:48 PM, Lisandro Dalcin dalcinl at gmail.com wrote:
On Tue, May 26, 2009 at 2:15 PM, Matthew Knepley knepley at gmail.com
wrote:
On Tue, May 26, 2009 at 12:07 PM, Lisandro Dalcin dalcinl at gmail.com
wrote:
The only real issue I see here is that the term diagonal
On Tue, May 26, 2009 at 2:10 PM, Jed Brown jed at 59a2.org wrote:
Matthew Knepley wrote:
Yes, I want to go into the code and every place we query the diagonal
block, replace cMap by rMap. This is the only thing that makes sense
to me. By my definitions, diagonal blocks are disjoint, so
We are about to release a new FMM code based upon PETSc. The paper is on
the arXiv and a link is on this website:
http://barbagroup.bu.edu/Barba_group/PetFMM.html
The mailing list is below. This first release will have only 2D, but I have
all the
3D stuff done except for the kernel expansions.
On Wed, May 27, 2009 at 2:50 PM, Barry Smith bsmith at mcs.anl.gov wrote:
On May 27, 2009, at 2:23 PM, Jed Brown wrote:
Gee, having vectors with a little extra space at the end is hard!
Notation: J is the Jacobian, P is the associated preconditioning matrix
PCFieldSplit gets the full row
Done.
Matt
On Thu, Apr 30, 2009 at 6:32 PM, Lisandro Dalcin dalcinl at gmail.com wrote:
Any good reason for petsc-dev/src/dm/ao/utils directory (and the
contained makefile) to be there? If not, could any of you remove it?
--
Lisandro Dalc?n
---
Centro Internacional de
On Thu, Jun 4, 2009 at 4:15 PM, Satish Balay balay at mcs.anl.gov wrote:
On Thu, 4 Jun 2009, Matthew Knepley wrote:
On Thu, Jun 4, 2009 at 3:37 PM, Satish Balay balay at mcs.anl.gov wrote:
Barry/Matt,
Should we make optfile in PetscIntialize(optfile) optional?
Why would
On Thu, Jun 4, 2009 at 4:51 PM, Satish Balay balay at mcs.anl.gov wrote:
On Thu, 4 Jun 2009, Matthew Knepley wrote:
Why would the user give an argument for optfile, but not intend it to
be
found?
Normally command line options is the way to pass in optional flags to
a petsc
I think there is too much setup that can be customized from the command
line. I
don't think you can decompose like that.
Matt
On Thu, Jun 4, 2009 at 5:33 PM, Lisandro Dalcin dalcinl at gmail.com wrote:
On Thu, Jun 4, 2009 at 6:17 PM, Satish Balay balay at mcs.anl.gov wrote:
This doesn't
Fixed.
Matt
On Fri, Jun 5, 2009 at 6:50 AM, Jed Brown jed at 59a2.org wrote:
plog.c:1557:
| ierr = PetscFPrintf(comm, fd, %20s %5d %5d %9d %g\n,
stageLog-classLog-classInfo[oclass].name,
| classInfo[oclass].creations,
classInfo[oclass].destructions,
In petscdmmod.F, I have
module petscdadef
use petscksp
#include finclude/petscda.h
#include finclude/ftn-custom/petscdadef.h90
end module
Why is KSP in there? This creates a circular dependence between libraries.
Looks like
it was introduced in 13712 (Satish). I am taking
I think its legitimate to only want some of our crap.
Matt
On Tue, Jun 9, 2009 at 4:16 PM, Barry Smith bsmith at mcs.anl.gov wrote:
In the old days we stored the compiled PETSc into a collection of
libraries to make linking faster.
Today, is there any reason not to just have all the
Also, I am not sure I buy the link time argument. Since your linker is
inspecting things
and only taking what it needs, shouldn't the search time be much reduced in
smaller
libraries?
Matt
On Tue, Jun 9, 2009 at 4:59 PM, Matthew Knepley knepley at gmail.com wrote:
I would ask: why
PETSc (at least what petsc4py
uses, that IMHO is a lot) is really tiny compared to any practical
run.
On Tue, Jun 9, 2009 at 4:55 PM, Jed Brown jed at 59a2.org wrote:
Barry Smith wrote:
On Jun 9, 2009, at 4:37 PM, Matthew Knepley wrote:
I think its legitimate to only want
, that IMHO is a lot) is really tiny compared to any practical
run.
On Tue, Jun 9, 2009 at 4:55 PM, Jed Brown jed at 59a2.org wrote:
Barry Smith wrote:
On Jun 9, 2009, at 4:37 PM, Matthew Knepley wrote:
I think its legitimate to only want some of our crap
If I understand correctly, the objection is that apply for MatShell is
apply(Mat, Vec, Vec)
and for PCShell is
apply(ctx, Vec, Vec)
I agree that this is crap. Its an easy fix to jsut pass the PC instead of
the ctx and let the user pull out the ctx manually if he wants. I am for
this.
On Wed, Jun 10, 2009 at 11:33 AM, Jed Brown jed at 59a2.org wrote:
Barry Smith wrote:
Go ahead. This has been on the todo list for a decade.
I see two ways to do it. One is to define a PCOperation enum and have
PCShellSetOperation (analogous to MatShell), the other is to just update
(like
Objective C). This already happens with ComposeFunction()
b) maintain type safety with replacement (I think ObjC does this, but we
don't)
c) have everything compile down to fast dispatch (I don't know if ObjC
does this, and it requires JIT)
Matt
On Jun 10, 2009, at 11:40 AM, Matthew
Are there ever any tests of this shit? It broke IMMEDIATELY on my box.
Matt
On Wed, Jun 10, 2009 at 11:02 PM, Satish Balay balay at mcs.anl.gov wrote:
On Wed, 10 Jun 2009, Barry Smith wrote:
All petsc-dev users. Next time you pull petsc-dev make sure you pull
config/BuildSystem also
Works now. However, the Scq options do not seem to be tested.
Matt
On Thu, Jun 11, 2009 at 6:51 AM, Matthew Knepley knepley at gmail.com wrote:
Are there ever any tests of this shit? It broke IMMEDIATELY on my box.
Matt
On Wed, Jun 10, 2009 at 11:02 PM, Satish Balay balay
On Thu, Jun 11, 2009 at 10:55 AM, Barry Smith bsmith at mcs.anl.gov wrote:
On Jun 11, 2009, at 7:27 AM, Matthew Knepley wrote:
Works now.
What was wrong with it?
Evidently, bad merge here.
However, the Scq options do not seem to be tested.
Correct. You can add a test
On Fri, Jun 12, 2009 at 11:52 PM, Satish Balay balay at mcs.anl.gov wrote:
On Fri, 12 Jun 2009, Matthew Knepley wrote:
I thought we have taken this support out, however it still seems to be
there. You can try it by adding
Why was mpe.py deleted? I see it deleted at
http
On Wed, Jan 14, 2009 at 3:15 PM, Barry Smith bsmith at mcs.anl.gov wrote:
From the PETSc developers guide
Do not use {\em if (rank == 0)} or {\em if (v == PETSC\_NULL)} or {\em if
(flg == PETSC\_TRUE)} or {\em if (flg == PETSC\_FALSE)}
instead use {\em if (!rank)} or {\em if (!v)} or
I have no problem with this convention. However, I just checked the C
standard,
and on page 254, in the stddef.h section, it says
NULL is an implementation specific null pointer constant.
So technically, it does not have to be 0.
Matt
On Wed, Jan 14, 2009 at 4:00 PM, Barry Smith bsmith at
Check out $PETSC_DIR/bin/configVars.py
Matt
On Thu, Jan 22, 2009 at 4:38 PM, Satish Balay balay at mcs.anl.gov wrote:
On Thu, 22 Jan 2009, Boyana Norris wrote:
Is there a script that can be used to query the values of various variables
defined in petscvariables makefile snippets (similar
is a separate project. It has its own repository
2) script is superfluous for getting the variables
Matt
Boyana
Matthew Knepley wrote:
On Thu, Jan 22, 2009 at 5:15 PM, Boyana Norris norris at mcs.anl.gov wrote:
Thank you both -- I'd prefer to use configVars.py, but it seems to depend
On Thu, Jan 22, 2009 at 6:06 PM, Boyana Norris norris at mcs.anl.gov wrote:
Matthew Knepley wrote:
On Thu, Jan 22, 2009 at 5:31 PM, Boyana Norris norris at mcs.anl.gov wrote:
Ah, ok, that explains it -- it's not in the installed version (the one
after
make install). Should it be? I
On Thu, Jan 22, 2009 at 6:18 PM, Barry Smith bsmith at mcs.anl.gov wrote:
On Jan 22, 2009, at 6:17 PM, Matthew Knepley wrote:
I meant something slightly different. I mean that the 'script' module
is really doing very
little work in this example. It is only processing command line
options
There is a problem with this. I don't really use 'script', but I do
use ArgDB, so we
still need BuildSystem.
Matt
On Thu, Jan 22, 2009 at 6:34 PM, Matthew Knepley knepley at gmail.com wrote:
I will split it into two pieces. One with a nice interface, that only
works in the
source
On Fri, Dec 4, 2009 at 8:52 AM, Barry Smith bsmith at mcs.anl.gov wrote:
On Dec 4, 2009, at 2:58 AM, Jed Brown wrote:
On Thu, 3 Dec 2009 16:40:10 -0600, Barry Smith bsmith at mcs.anl.gov
wrote:
The vector has a pointer to the DM so the VecView() for that
derived vector class has access
On Fri, Dec 4, 2009 at 10:42 PM, Barry Smith bsmith at mcs.anl.gov wrote:
Suggestion:
1) Discard PETSc
2) Develop a general Py{CL, CUDA, OpenMP-C} system that dispatches tasks
onto GPUs and multi-core systems (generally we would have one python process
per compute node and local
On Sat, Dec 5, 2009 at 1:06 PM, Jed Brown jed at 59a2.org wrote:
This is an interesting proposal. Two thoughts:
Residual and Jacobian evaluation cannot be written in Python (though it
can be prototyped there). After a discretization is chosen, the physics
is usually representable as a tiny
and C++ compared to Pyrex.
Brad
Matthew Knepley wrote:
On Fri, Dec 4, 2009 at 10:42 PM, Barry Smith bsmith at mcs.anl.gov
mailto:bsmith at mcs.anl.gov wrote:
Suggestion:
1) Discard PETSc
2) Develop a general Py{CL, CUDA, OpenMP-C} system that dispatches
On Sat, Dec 5, 2009 at 3:50 PM, Jed Brown jed at 59a2.org wrote:
On Fri, 4 Dec 2009 22:42:35 -0600, Barry Smith bsmith at mcs.anl.gov wrote:
generally we would have one python process per compute node and local
parallelism would be done via the low-level kernels to the cores
and/or GPUs.
On Sat, Dec 5, 2009 at 4:25 PM, Jed Brown jed at 59a2.org wrote:
On Sat, 5 Dec 2009 16:02:38 -0600, Matthew Knepley knepley at gmail.com
wrote:
I need to understand better. You are asking about the case where we have
many GPUs and one CPU? If its always one or two GPUs per CPU I do
On Sat, Dec 5, 2009 at 6:01 PM, Jed Brown jed at 59a2.org wrote:
On Sat, 5 Dec 2009 16:50:38 -0600, Matthew Knepley knepley at gmail.com
wrote:
You assign a few threads per element to calculate the FEM
integral. You could maintain this unassembled if you only need
actions.
You can also
kernels for your inner
loops.
Matt
Chris
On Dec 6, 2009, at 9:43 AM, Jed Brown wrote:
On Sat, 5 Dec 2009 14:23:45 -0600, Matthew Knepley knepley at gmail.com
wrote:
We do impose some of our process on users already, such as CHKERRQ and
PetscMalloc.
CHKERRQ is still optional, it's
On Wed, Dec 9, 2009 at 6:24 PM, Barry Smith bsmith at mcs.anl.gov wrote:
in compilers.py it logs and error message but keeps on running if the C
libraries cannot be used from Fortran. Why is this? It seems totally insane,
it just delays the error until later when it is confusing and unclear
On Wed, Dec 9, 2009 at 6:33 PM, Barry Smith bsmith at mcs.anl.gov wrote:
On Dec 9, 2009, at 6:29 PM, Matthew Knepley wrote:
On Wed, Dec 9, 2009 at 6:24 PM, Barry Smith bsmith at mcs.anl.gov wrote:
in compilers.py it logs and error message but keeps on running if the C
libraries cannot
On Thu, Dec 10, 2009 at 3:03 PM, Jed Brown jed at 59a2.org wrote:
Can KSPView before KSPSolve reasonably be made to work?
What breaks? If you mean that it is not broken, but information is different
than it
is after KSPSetUp(), I would say no.
matt
-- Forwarded message --
Not sure if you would need API additions. Can't everything you describe be
handled through defining the grid, operator, residual, and projections?
Matt
On Sun, Dec 13, 2009 at 8:59 AM, Jed Brown jed at 59a2.org wrote:
I can think of a few ways to implement the following multilevel scheme,
On Sun, Dec 13, 2009 at 5:56 PM, Jed Brown jed at 59a2.org wrote:
On Sun, 13 Dec 2009 16:59:39 -0600, Barry Smith bsmith at mcs.anl.gov
wrote:
It was never fully developed (i.e. things might not work). It
could be we should switch to using that one and slowly remove the
regular refine
On Sun, Dec 13, 2009 at 6:41 PM, Jed Brown jed at 59a2.org wrote:
On Sun, 13 Dec 2009 18:06:04 -0600, Matthew Knepley knepley at gmail.com
wrote:
I will explain the history since I am to blame for this one. I needed
coarsenHierarchy because repeated coarsening can cause degradation
On Sun, Dec 13, 2009 at 9:36 PM, Barry Smith bsmith at mcs.anl.gov wrote:
So if I pass --with-mpi-shared into MPI.py it TESTS if the MPI libraries
are shared. Fine. But if it is download-mpich IT DOES NOT try to make shared
libraries for MPICH. It only tries to make shared libraries for MPI
On Mon, Dec 14, 2009 at 8:28 AM, Barry Smith bsmith at mcs.anl.gov wrote:
On Dec 14, 2009, at 8:23 AM, Matthew Knepley wrote:
On Mon, Dec 14, 2009 at 8:21 AM, Barry Smith bsmith at mcs.anl.gov wrote:
On Dec 14, 2009, at 8:16 AM, Matthew Knepley wrote:
I strongly disagree
On Mon, Dec 14, 2009 at 10:28 AM, Lisandro Dalc?n dalcinl at gmail.com wrote:
On Mon, Dec 14, 2009 at 11:34 AM, Matthew Knepley knepley at gmail.com
wrote:
If we are using shared libraries, and the MPI shared library was broken,
the
configure failed. This option was added
to allow
On Mon, Dec 14, 2009 at 10:39 AM, Lisandro Dalc?n dalcinl at gmail.com wrote:
On Mon, Dec 14, 2009 at 1:33 PM, Matthew Knepley knepley at gmail.com
wrote:
On Mon, Dec 14, 2009 at 10:28 AM, Lisandro Dalc?n dalcinl at gmail.com
wrote:
On Mon, Dec 14, 2009 at 11:34 AM, Matthew Knepley
there is a stack smash and no other info. This is completely fucking
my development right now.
Matt
Satish
On Mon, 21 Dec 2009, Matthew Knepley wrote:
I killed my directory, removed sowing from extpacks, and reran configure,
but I get
TESTING
-has_star = (nstar 0);
-arg-name = (char *)MALLOC( strlen(p) + 1 );
+arg-name = (char *)MALLOC( strlen(p) + 10 );
strcpy( arg-name, p );
/* We can't output the name just yet, because if it is
On Mon, 21 Dec 2009, Matthew Knepley wrote:
The problem appears
On Fri, Dec 25, 2009 at 10:14 AM, Satish Balay balay at mcs.anl.gov wrote:
Did my suggested change not work for you?
No. It did not even fix the stack smash. I fixed that, but then it SEGV
somewhere else.
Matt
Satish
On Thu, 24 Dec 2009, Matthew Knepley wrote:
I spent a bunch
PETSC_HAVE_FORTRAN_CAPS -fuscore PETSC_HAVE_FORTRAN_UNDERSCORE matrix.c
Satish
On Fri, 25 Dec 2009, Matthew Knepley wrote:
Valgrind is not clean for me with the change.
Matt
On Fri, Dec 25, 2009 at 10:28 AM, Satish Balay balay at mcs.anl.gov
wrote:
Well - normally
etc..]
Satish
On Fri, 25 Dec 2009, Matthew Knepley wrote:
Here is the valgrind for your +100 fix:
knepley at khan:/PETSc3/petsc/petsc-dev/src/mat/utils$ valgrind
/PETSc3/petsc/petsc-dev/linux-gnu-cxx-debug/bin/bfort -dir
/PETSc3/petsc/petsc-dev/src/mat/utils/ftn-auto -mnative -ansi
On Tue, Apr 7, 2009 at 9:51 PM, Ahmed El Zein ahmed at azein.com wrote:
On Tue, 2009-04-07 at 10:39 -0500, Matthew Knepley wrote:
On Tue, Apr 7, 2009 at 9:07 AM, Stephen Ball
Stephen.R.Ball at awe.co.uk wrote:
Hi
We are keen to start investigating whether or not PETSc
PetscPopSignalHandler()
Matt
On Thu, Apr 16, 2009 at 2:44 PM, KACVINSKY Tom Tom.KACVINSKY at 3ds.comwrote:
I want to completely disable the signal handling in PETSc. We have our
own
traceback generator and PETSc's signal handler interferes with this. What
is
the best way of turning
On Thu, Apr 30, 2009 at 7:14 AM, Jed Brown jed at 59a2.org wrote:
Matthew Knepley wrote:
We should just treat LocalForm as GetArray, and increase the state on
RestoreLocalForm.
I thought of this, but like the maximum better. With GetArray, you
don't have a nice way to determine whether
On Thu, Apr 30, 2009 at 8:12 AM, Jed Brown jed at 59a2.org wrote:
Matthew Knepley wrote:
I misunderstood. I thought you were checking for element maxima, which
would tell you if something changed. You are just checking state
numbers?
Yes.
I did not think these had any guarantees
On Thu, Apr 30, 2009 at 9:08 AM, Jed Brown jed at 59a2.org wrote:
Matthew Knepley wrote:
I am opposed to this, because I really do not like the state mechanism
in the first place. It is dangerous (as your debugging shows), and
relying on it makes it very easy for people to introduce
I have seen this on Windows when we get random file protection errors.
MAtt
On Tue, Mar 31, 2009 at 9:09 PM, Lisandro Dalcin dalcinl at gmail.com wrote:
After a configure make, make test failed because of broken libs.
I've noticed this:
$ grep 'ar:' cygwin-c-debug/conf/make.log
On Wed, Sep 17, 2008 at 10:51 PM, Ahmed El Zein ahmed at azein.com wrote:
On Wed, 2008-09-17 at 21:33 -0500, Barry Smith wrote:
Ahmed,
This is very cool.
On Sep 17, 2008, at 8:05 PM, Ahmed El Zein wrote:
Harald,
I am working on implementing SpMV on a NVIDIA GPU with CUDA for SML
On Thu, Sep 18, 2008 at 1:34 PM, Ahmed El Zein ahmed at azein.com wrote:
On Thu, 2008-09-18 at 10:20 -0500, Matthew Knepley wrote:
A question that I had regarding the PETSc code when I was thinking
about
this was:
You have the SeqAIJ matrix type and the the MPIAIJ type built around
I can also debug just fine or I would scream bloody murder.
Matt
On Thu, May 1, 2008 at 10:23 AM, Satish Balay balay at mcs.anl.gov wrote:
What machine is this?
Its been working forever for me on linux.
Satish
---
On Thu, May 8, 2008 at 10:09 PM, Dave May dave.mayhem23 at gmail.com wrote:
Hey petsc folk,
What is the best way to obtain timing information to profile the
performance
of KSPSolve (or SNESSolve)? Currently I have written some specific KSP
monitors,
but it I think it would be useful to
it seems to put redundant
info into the
KSP, and it woul dbe meaningless if logging is turned off. Maybe there
should be a
history mechanism in the logging. Right now it is completely based
upon aggregation.
Matt
Cheers,
Dave
On Fri, May 9, 2008 at 1:19 PM, Matthew Knepley knepley
2008/5/10 Lisandro Dalcin dalcinl at gmail.com:
I've noticed that in 2.3.3 all PETSc cookies are initialized to zero
in sources, but in DEV they are not initialized. What's the reason of
this change?
Not sure, but I would guess some compiler probably did not like it.
Matt
--
701 - 800 of 2694 matches
Mail list logo